The stock structure of the valuable tropical shad Tenualosa ilisha ('hilsa') has been studied in Bangladesh and India by analysing morphometric and genetic data. However, these studies had a narrow geographic scope and their results conflict. We made a comprehensive study of the stock structure of hilsa with otolith microchemistry in conjunction with complementary genetic and morphometric studies of the same fish. We examined the trace-element composition of the otolith cores of hilsa with laser-ablation inductively coupled plasma mass spectrometry. The otoliths of fish from 19 collections at 13 sites in Bangladesh and 6 collections at 4 sites from elsewhere within the species' range (Kuwait, SE India, Myanmar and Sumatra) were analysed for 8 trace elements. Samples were collected from Bangladesh mainly during 2 comprehensive surveys (1996 and 1997). When these data were analysed separately, there were significant differences in otolith composition among sites. However, when both years' data were analysed together, there were few significant differences among sites, and some sites separated by hundreds of kilometres that were sampled in different seasons and years had very similar compositions. This was in spite of both large seasonal intra-site and between-site differences in water chemistry. Repeat samples from 5 sites (4 in Bangladesh) showed that differences in otolith composition at a single site were significant and of similar magnitude to that found among sites. Our results support the conclusion from allozyme studies that there is extensive movement and mixing of hilsa throughout Bangladesh, and therefore the population should be managed as a single stock. Genetic and otolith data both showed that hilsa from SE India and Myanmar were not significantly different from fish collected in coastal areas of Bangladesh, and suggest that hilsa in the Bay of Bengal were a single stock. Both methods also separated fish from Sumatra and Kuwait from other sites, providing strong evidence of separate stocks in those regions. In contrast, morphometric studies separate fish from several nearby sites in Bangladesh, but these differences are likely to be largely due to phenotypic variability and are unlikely to be geneticallybased. Our results suggest that otolith microchemistry may be a good proxy for genetic structure at large scales where differences in water chemistry are highly likely. However, for sedentary species and those without distinct spawning and non-breeding areas, it requires both comprehensive and repeated sampling at finer scales before any confidence should be placed in the results.